A Component Reusability Analysis Method Integrating Semantic Similarity and Reuse Value

Evaluating the reusability of existing components based on functional descriptions is critical for component-level software reuse. Current approaches have limitations such as underutilization of multimodal data, neglect of components’ inherent reuse value, and large model parameters. To address these challenges, this paper presents a reusable component analysis method (SSRV-RCA) that combines semantic similarity and reuse value. The method comprises a lightweight multimodal contrastive learning code representation model (MCL-CodeRep) and a component reuse value evaluation system (CRVMS). MCLCodeRep bridges the semantic gap between structured code and natural language using cross-modal contrastive learning to calculate semantic similarity between functional descriptions and components. CRVMS quantifies component reuse value across three dimensions-code quality, functional characteristics, and interface design standards-using the Analytic Hierarchy Process (AHP). By integrating semantic similarity and reuse value through weighted combination, a reusability indicator is generated. Experimental results demonstrate that MCL-CodeRep performs effectively in code search and clone detection tasks. Compared to manual expert evaluations, SSRV-RCA exhibits high accuracy in practical applications, effectively identifying reusable components and strongly supporting efficient component reuse in software development.